The Transformation Mindset: Bridging the 20% Adoption Gap with Tim Flagg, UKAI

John Emmerson: Hi, I’m here with Tim Flagg, CEO and co-founder of UKAI, the UK’s dedicated trade association for AI businesses. Towards the end of last year, you did some research with SMEs that flagged a £198 billion untapped AI opportunity for UK SMEs. Where do you feel most mid-market firms actually are with AI adoption right now, and what’s holding back the ones that haven’t moved yet?

Tim Flagg: Great to be here, John. It’s a really exciting time because people are not just talking about AI now — they’re actually building it and thinking seriously about adoption. There’s still a lot of anxiety: about what AI is, how it’s going to affect people individually, and the anxiety about how to get started. I want to look at this through a very human lens.

Tim Flagg: In terms of how businesses are actually adopting this: the figures seem to be quite consistent. Something like 20 to 25 per cent of businesses are now in that exploratory phase where they’re trying things, getting started. That’s good. But the question is what’s happening with the other 75 per cent. Are they just doing nothing? Or is it the use of shadow AI — where a lot of people are using completely different systems, bringing their own platforms and tools, not adhering to their organisation’s guardrails? There is a big challenge there around security.

Tim Flagg: There’s also a challenge within the organisation itself. What happens if everyone’s not on the same page is that you try these pilots, you get so far and then they fail, and because everyone’s on different pages and using different systems, people say: well, AI failed. Without really thinking about which bit of AI failed and whether it was a fair test. Unless we’re doing this in a systemic way, we risk undermining the opportunity to actually drive adoption and get companies to scale.

John Emmerson: Within that 20 to 25 per cent who are adopting, what are the ones doing it best actually doing differently?

Tim Flagg: That’s the multi-million dollar question. Over the last 15 years there’s been a lot of talk about digital transformation. The key thing we’ve been talking about in transformation is mindset and culture change. And for me, this is as relevant as ever.

Tim Flagg: What do we mean by that? It’s about building the ability to be curious. Things are moving so quickly. There’s no way you can just learn about something and treat it as a one-and-done. You have to continually improve and learn about what’s coming next, scan for who the next vendor or tool is going to be. Unless you have that curiosity to go out there and explore, try things, read those blogs, listen to those podcasts, you’re just not going to be able to keep up.

Tim Flagg: The second part is resilience. With all of that scale and all those options happening at once, it can be overwhelming. And frankly, I think part of being in this industry is accepting that it is overwhelming. You can’t expect to understand everything. So you need the resilience to keep on learning, test things out, know that you’re going to test some things that will fail, and learn from them. That test-and-learn iterative approach, combined with data-driven decision-making, makes up what I call an adaptive mindset.

Tim Flagg: The whole successful AI transformation comes down to organisations that have started to build that culture and that mindset. The companies who got into transformation ten years ago — for them, when AI came along, it was just another technology. They didn’t have to scramble. They’d already been scanning, already been testing, they had that curiosity and resilience built in. And so they just brought AI into their normal daily routines and scaled it up from there. That’s a very small minority. What we’re seeing now is the rest of the companies almost running around saying: we need to do an AI transformation, and specifically we need to get AI into the organisation — without thinking about the culture, without thinking about the business need, without thinking about what the technology would actually allow them to do.

John Emmerson: How important is the people and HR function in driving AI adoption?

Tim Flagg: You’re absolutely right that it does come down to the people. That’s both the opportunity and the challenge. At a senior level, leaders need to understand their role in setting out the vision and encouraging that culture. At the managerial level, it’s about understanding how to encourage your team and set realistic KPIs that they can actually work towards. And what we’re starting to see is that managers are having to think about what hybrid teams look like — how do they encourage their teams to use not just AI-powered tools but also agents, and where do those agents fit into the team? What does that do to the KPIs and responsibilities of the humans?

Tim Flagg: Those managers have a really important role to play — putting human judgement into those agentic systems, and then providing quality control at the other end. A real change in the role of managers. At the everyday employee level, it’s about giving people the confidence to experiment and try new things. What we see at that level is that it’s often just about confidence. People feel they don’t know what AI is, because it can be such a bewildering thing. But once you start demystifying it and showing them simpler ways of getting started — ways that are actually relevant to their jobs — they love it. You go from an audience that’s leaning back, arms crossed, not seeing how it’s relevant, to suddenly leaning forward going: wow.

John Emmerson: Top down or bottom up for AI adoption — what’s your view?

Tim Flagg: In some ways it’s both, but let me zoom out to a national level. It’s all about building confidence and trust. If the humans in this country don’t understand the point of using AI and they’re not using it, then they’re not going to be the consumers buying AI-powered products and services, or the employees engaging with AI-powered tools, or the citizens using AI-powered tools that councils and government are trying to deploy. So we need to think of it from that perspective.

Tim Flagg: The best way we can do that is through education and trust. Education: help people understand what AI is — not just chatbots, which almost everyone’s heard of now, but the broader applications of machine learning, analytics, data science. Give them examples of where AI is delivering for good, in healthcare or in everyday tasks. And ultimately, once they understand it, they will trust it. So there’s a role to play nationally in driving consumer demand — and of course that’s alongside the top-down leadership role.

John Emmerson: You touched on data as a real starting point. For businesses with fragmented CRM, messy ERP data, legacy systems — what’s the minimum viable data foundation before AI can do anything useful?

Tim Flagg: The legacy data issue is massive. Particularly when companies have merged, gone from one CRM to another, or bought in datasets. Those datasets often just sit there, locked away. There is some use you can make of data as it is, but the real opportunity is using AI-powered tools to go in and actually clean that data up — recodify it, make it interoperable across multiple different tools. We’re starting to see a number of our members now operating exactly in this space, going into large organisations and building consistent, interoperable data sets that can then form the bedrock for the next AI-driven project those companies want to do.

Tim Flagg: This comes back to what we were talking about earlier. Often companies will want to race because they feel like they’re behind the curve and they’ve got a bit of FOMO. The top-down question comes: how do we make an AI for our business? That’s often the wrong question. What they should be thinking is: what’s the business need? What’s the problem? And then how do we design some technology that actually addresses that problem? Starting with cleaning up the data is sometimes the less exciting part of AI transformation, but it’s a fundamental part. Once you’ve done that, it unlocks all of the greater things you can do as well.

John Emmerson: You advise government, you consult with ministers including the AI minister. What are you hearing from policymakers that mid-market businesses need to hear?

Tim Flagg: I was chatting to the AI minister, Kanishka Narayan, a couple of nights ago. He is very well versed in AI — he comes from a VC and technology background and spends a lot of time talking not just to our members but to businesses across the country. What we’re hearing from government in terms of how medium-sized businesses can actually help is through two things. One is adoption. Something like 73 per cent of employees in the UK are in a small or medium-sized enterprise, so SMEs have a particularly important role to play in being the champions, driving the cultures we were talking about, and bringing their employees — and the consumers who buy their products and services — with them.

Tim Flagg: The other part is that these businesses can provide the services and the case studies for how AI can deliver solutions to central government, local government and other businesses as well. The challenge there comes down to procurement and investment. The government holds within its remit the ability to fast-track procurements and make it easier through agencies like UKRI. There’s also the Sovereign AI unit now with a £500 million fund. The bit we’re trying to make happen is making sure businesses that have the solutions can actually meet the demand from those departments — and even though it seems simple, it is actually quite a hard thing to bring those two groups together.

John Emmerson: Mid-market firms are getting pitched AI solutions constantly. What should a buyer actually be asking vendors that most of them don’t think to ask?

Tim Flagg: The challenge for the people building these businesses often comes down to what IP they’re actually developing. As a buyer, you want to know that whoever you’re working with actually has the right mix: something they’ve developed themselves and own the IP for, but also something which is robust. That’s going to be a critical thing to look out for.

Tim Flagg: The other really big area is understanding the ethics of the vendor you’re working with. Think back 15 to 20 years ago, when ESG emerged and became a fundamental part of procurement. We’re going to see that with AI businesses as well, where buyers are looking to ensure providers have sourced their data in an ethical and responsible way, are thinking about sustainability, are looking at bias within their systems. These aren’t fluffy considerations. They’re commercial ones. And a lot of responsible AI companies are already recognising that and identifying themselves accordingly.

John Emmerson: There’s been a lot of talk about the Wild West atmosphere in parts of the AI vendor ecosystem. Some mid-market leaders worry that governance, trust and compliance risk is slowing their adoption down. Is that real?

Tim Flagg: There is a bit of paralysis. Towards the end of last year, there was a lot of uncertainty around the proposed UK AI Bill, and people didn’t really know what was going to be in it. In the end it was quietly dropped because there was just too much it could cover. But that uncertainty actually meant a lot of organisations held off making decisions.

Tim Flagg: The critical shift in mindset we need is to see regulation not as something that limits you, but as something that sets standards and reassures people. The UK is really good at standard-setting. We’ve set standards for the world across many different industries. Within AI, we can set standards for responsible AI, ethical AI, Green AI — making the whole AI stack more sustainable in how it uses energy and water. Going back to the Wild West comment: that was absolutely how it felt a year and a half ago when we were setting up UKAI. And it was one of the reasons we wanted to do this as business owners. We came together and that became the foundation for UKAI — because we realised that moving from the Wild West to a mature industry that can actually add value to society and build trust is only possible if the industry itself takes responsibility for the negative sides.

John Emmerson: UKAI itself has gone through its own AI adoption journey with Ready Intelligence. What did you learn from it?

Tim Flagg: It really brought into focus the importance of testing things out and failing sometimes. With a lot of these tools, there’s no neat guidebook that says: here’s how you start, step one, two and three. You have to just try it out. So the main learning is: we bolted a bunch of things together, we played around with things, we saw what works. And then there’s the human element — getting things to work technically is one thing, getting them to work in a way that suits your team and the individuals in it is the other thing. What we’re seeing more of now is that we can build in these different component parts and really customise things around our actual workflow. That’s what we’re continuing to develop.

What is the £198 billion AI opportunity for UK SMEs?

UKAI’s research identified a £198 billion untapped AI opportunity for UK SMEs, based on the productivity gains available to small and medium-sized enterprises that adopt AI at scale. The figure reflects the gap between current adoption rates and what is achievable if businesses move from exploratory use to systemic implementation. Tim Flagg notes that something like 73 per cent of UK employees work in SMEs, making mid-market adoption a critical lever for national economic growth.

What is shadow AI and why does it matter for mid-market businesses?

Shadow AI refers to the use of AI tools by employees outside the guardrails, policies and approved systems that an organisation has put in place. Rather than using the AI platforms their organisation has sanctioned, employees bring their own tools and platforms — creating significant security risks and making it impossible to achieve systemic adoption. Tim Flagg argues that shadow AI is a symptom of a missing corporate AI strategy: if leaders don’t provide the tools and the structure, teams will find their own — and those tools will not be secure or compliant.

Why do most AI transformation efforts fail to get past the pilot stage?

Tim Flagg’s view is that most AI transformation efforts fail because organisations approach them as a technology project rather than a cultural one. They race to get AI into the organisation without thinking about the culture, the business need or the problem they’re actually trying to solve. Without a shared adaptive mindset — curiosity, resilience and data-driven experimentation — across the organisation, pilots succeed in isolation and then stall when they try to scale.

What is an adaptive mindset and why does it matter for AI adoption?

Tim Flagg defines an adaptive mindset as the combination of curiosity, resilience and data-driven experimentation that characterises organisations successfully scaling AI. Curiosity means continuously scanning for what is new and trying things rather than waiting for a playbook. Resilience means accepting that not everything will work and learning from what fails. Data-driven experimentation means measuring what you do and using that data to make better decisions. Companies that built this culture during their digital transformation programmes ten years ago found AI adoption far easier, because it was just another technology to absorb into an already adaptive way of working.

What should mid-market businesses do about legacy data before investing in AI?

Tim Flagg’s advice is to use AI to fix your data before you try to build on it. Legacy data — fragmented across old CRMs, merged databases and bought-in datasets — is one of the most common blockers to AI value. A growing number of specialist businesses now offer AI-powered data clean-up services that recodify legacy data and make it interoperable across multiple systems. Organisations that start by asking ‘how do we make an AI for our business?’ are starting from the wrong place; the right starting point is always the business problem.

What should mid-market businesses ask AI vendors before they buy?

Tim Flagg identifies two critical questions. First, what IP does the vendor actually own? With vibe coding making it easier than ever to build AI products quickly, buyers need to understand whether a vendor has built something proprietary and defensible or assembled something that could be replicated by anyone. Second, what are the vendor’s ethics? Just as ESG credentials became a standard procurement consideration over the past 20 years, responsible AI credentials are becoming commercial necessities — covering data sourcing, bias mitigation, sustainability and transparency.

Is UK AI regulation helping or hindering business adoption?

Tim Flagg’s view is that regulation, properly framed, should be seen as standard-setting rather than restriction. The uncertainty around the proposed UK AI Bill towards the end of 2025 created paralysis in some organisations. The Bill was subsequently dropped. Tim’s argument is that the UK has a strong track record of setting global standards across industries, and the opportunity within AI is to do the same: set standards for responsible AI, ethical AI and Green AI that build trust with consumers, employees and trading partners, and that allow the industry to grow.